CN115331096B - Mining subsidence area identification method, system, storage medium and electronic equipment - Google Patents
Mining subsidence area identification method, system, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN115331096B CN115331096B CN202210749586.6A CN202210749586A CN115331096B CN 115331096 B CN115331096 B CN 115331096B CN 202210749586 A CN202210749586 A CN 202210749586A CN 115331096 B CN115331096 B CN 115331096B
- Authority
- CN
- China
- Prior art keywords
- original
- area
- mining subsidence
- module
- phase
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Abstract
The invention relates to a mining subsidence area identification method, a mining subsidence area identification system, a storage medium and electronic equipment, wherein the mining subsidence area identification method comprises the following steps: generating a differential interference pattern of each original monitoring area by using a two-rail method DINSAR technology; filtering and converting each differential interference pattern to obtain each phase pattern; marking mining subsidence areas and background areas in each phase map, obtaining and cutting the marked phase map, and obtaining a plurality of cutting phase sample maps; training the improved U-net model based on all the clipping phase sample patterns to obtain and utilize a target network model to identify a target differential interference pattern of the area to be monitored, and obtaining identification results of a mining subsidence area and a background area in the area to be monitored. According to the invention, the intelligent recognition is carried out on the DINSAR interference phase diagram by improving the U-Net model, so that the semantic segmentation precision of the U-Net model is improved, and the recognition capability of subsidence areas with different scales is improved.
Description
Technical Field
The present invention relates to the field of image recognition technologies, and in particular, to a mining subsidence area recognition method, system, storage medium, and electronic device.
Background
The subsidence of the earth surface caused by the exploitation of mineral resources has destructive effects on the construction, traffic, land utilization and ecological environment, and can cause economic loss of government, and the uncertainty of the subsidence time can endanger the life health and safety of people. In addition, extraction of mining subsidence areas has an indicative effect in areas where there is illegal mining of the subsurface. The intelligent recognition of the subsidence range of the mining area by combining deep learning with the DINSAR technology is a novel method for monitoring mining subsidence, and can provide effective help for geological disaster risk assessment, land utilization management planning, mine area safety guarantee and the like.
Remote sensing is a means of remote earth detection, can acquire image information in a large scale range and has strong timeliness. Radar remote sensing has all-weather working and cloud penetrating and fog penetrating capabilities throughout the day, and a synthetic aperture interferometry (InSAR) technology is considered to be the only technology at present capable of realizing high-precision ground surface height change monitoring in a large range. Compared with the traditional method for monitoring the surface deformation in field operation, the DINSAR technology can obtain the surface shape data, has the advantages of large monitoring range, higher spatial resolution, quick monitoring processing flow, no contact, no need of spending a large amount of manpower and material resources and no potential safety hazard. While the DInSAR technology is now quite mature, the time required to delineate the mining subsidence range using manual interpretation of the DInSAR monitoring results is long when dealing with the rapid response needs.
Early remote sensing classification methods are mainly based on the colors, shapes and textures of images to classify bottom layer features or middle layer features, but the methods need to manually extract features, have large workload and are not efficient enough, and are difficult to achieve better classification accuracy. Therefore, there is a need to propose a technical solution for efficiently and effectively identifying mining subsidence areas.
Disclosure of Invention
In order to solve the technical problems, the invention provides a mining subsidence area identification method, a mining subsidence area identification system, a storage medium and electronic equipment.
The technical scheme of the mining subsidence area identification method is as follows:
s1, generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated;
s2, filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain original phase patterns until the original phase patterns corresponding to each original differential interference pattern are obtained;
s3, marking mining subsidence areas and background areas in any original phase diagram, obtaining and cutting the marked phase diagram to obtain a plurality of cutting phase sample diagrams until a plurality of cutting phase sample diagrams of each original phase diagram are obtained;
s4, training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model;
and S5, identifying a target differential interferogram corresponding to the area to be monitored by utilizing the target network model, and obtaining identification results of the mining subsidence area and the background area in the area to be monitored.
The mining subsidence area identification method has the beneficial effects that:
according to the method, the improved U-Net model is used for intelligently identifying the DINSAR interference phase diagram, so that the semantic segmentation precision of the U-Net model is improved, and meanwhile, the identification capability of subsidence areas with different scales is improved.
On the basis of the scheme, the mining subsidence area identification method can be improved as follows.
Further, the generating, by using the two-track DInSAR technique, an original differential interferogram of any original monitoring area based on a first image and a second image of the any original monitoring area at different moments acquired from the same incident angle includes:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
Further, the construction process of the improved U-net model comprises the following steps:
and respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model.
Further, the clipping phase sample map includes: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
before the step S5, the method further includes:
s051, training the original U-net model based on all training sample graphs to obtain an original network model;
s052, sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original model precision of the original network model based on an average cross-correlation formula; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
s053, respectively inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
s054, judging whether the target model precision is larger than the original model precision, obtaining a judging result, and executing step S5 when the judging result is yes.
Further, the method further comprises the following steps: and when the judging result is negative, re-acquiring an original differential interference diagram of a new original monitoring area until the target model precision is greater than the original model precision, and executing step S5.
The technical scheme of the mining subsidence area identification system is as follows:
comprising the following steps: the device comprises a preprocessing module, a first processing module, a second processing module, a training module and an identification module;
the preprocessing module is used for: generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated;
the first processing module is used for: filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain an original phase pattern until the original phase pattern corresponding to each original differential interference pattern is obtained;
the second processing module is used for: marking mining subsidence areas and background areas in any original phase map, obtaining and cutting the marked phase map to obtain a plurality of cutting phase sample maps until a plurality of cutting phase sample maps of each original phase map are obtained;
the training module is used for: training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model;
the identification module is used for: and identifying a target differential interference pattern corresponding to the area to be monitored by utilizing the target network model to obtain identification results of the mining subsidence area and the background area in the area to be monitored.
The mining subsidence area identification system has the following beneficial effects:
the system provided by the invention can be used for intelligently identifying the DINSAR interference phase diagram through the improved U-Net model, so that the semantic segmentation precision of the U-Net model is improved, and the identification capability of subsidence areas with different scales is improved.
On the basis of the scheme, the mining subsidence area identification system can be improved as follows.
Further, the generating, by using the two-track DInSAR technique, an original differential interferogram of any original monitoring area based on a first image and a second image of the any original monitoring area at different moments acquired from the same incident angle includes:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
Further, the construction process of the improved U-net model comprises the following steps:
and respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model.
Further, the clipping phase sample map includes: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
before the identification module, further comprising: the device comprises a first verification module, a second verification module, a third verification module and a fourth verification module;
the first verification module is used for: training the original U-net model based on all training sample graphs to obtain an original network model;
the second verification module is used for: sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original model precision of the original network model based on an average cross-correlation formula; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
the third verification module is used for: inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
the fourth verification module is used for: and judging whether the precision of the target model is greater than that of the original model or not, obtaining a judging result, and calling the identification module when the judging result is yes.
The technical scheme of the storage medium is as follows:
the storage medium has stored therein instructions which, when read by a computer, cause the computer to perform the steps of a mining subsidence area identification method according to the present invention.
The technical scheme of the electronic equipment is as follows:
comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the computer program, causes the computer to perform the steps of a mining subsidence area identification method according to the invention.
Drawings
FIG. 1 is a flow chart of a mining subsidence area identification method according to an embodiment of the present invention;
FIG. 2 is a schematic illustration of an original monitored zone in a mining subsidence area identification method according to an embodiment of the present invention;
FIG. 3 is a differential interference pattern and a conversion processing effect diagram thereof in a mining subsidence area identification method according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating a method for identifying mining subsidence areas according to an embodiment of the present invention;
FIG. 5 is a block diagram of ECA modules in a mining subsidence area identification method according to an embodiment of the present invention;
FIG. 6 is a diagram of an original U-Net model structure in a mining subsidence area identification method according to an embodiment of the present invention;
FIG. 7 is a diagram of an improved U-Net model structure in a mining subsidence area identification method according to an embodiment of the present invention;
fig. 8 is a schematic structural view of a mining subsidence area identification system according to an embodiment of the present invention.
Detailed Description
As shown in fig. 1, a mining subsidence area identification method according to an embodiment of the present invention includes the following steps:
s1, generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated.
The number of the selected original monitoring areas in this embodiment is two, and specifically the selected original monitoring areas are: as shown in fig. 2, track numbers path113_frame116 and track numbers path113_frame121 correspond to the areas.
Wherein, the two-rail method DINSAR technology is the prior art. The D-InSAR technology repeatedly images the same monitoring area at a certain time interval and at the same incidence angle, wherein one of the two images is an interference image (a first image) before deformation, the other image is an interference image (a second image) obtained after deformation, the surface deformation can cause the distance change of radar line of sight (LOS), the distance change of the radar line of sight (LOS) is recorded in an interference phase, the interference phase is analyzed, and the surface deformation information is obtained through differential processing (removing the influence of reference ellipsoidal phase contribution and topography fluctuation on the interference phase). Therefore, after differential interference processing is carried out on the first image and the second image, a corresponding original differential interference pattern is obtained.
Specifically, for example, the selected data time span is 10 months 11 days to 12 months 22 days in 2019. The satellite return period for sentinel number 1 is 12 days. The two-track method D-InSAR can select the first image and the second image of the area range of Path113_frame116 from 10/11/2019/10/23, and perform differential interference processing on the first image and the second image, wherein any one of the first image and the second image can be used as a main image, and no limitation is set here.
S2, filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain an original phase map until the original phase map corresponding to each original differential interference pattern is obtained.
The process of filtering the original differential interference pattern to obtain the filtered differential interference pattern is the prior art, and is not repeated here.
Wherein, the original phase diagram is: the filtered differential interferogram is directly converted by command in commercial software such as GAMMA according to a formula. Specifically, as shown in fig. 3, the original phase diagram is a value [ -pi, pi) phase diagram. For example, assume that S1 and S2 represent measurement signals of the same target T on the ground at similar angles of incidence for SAR at position 1 and position 2, respectively, according to complex signal expression form:
wherein phi represents the phase, a 1 For the first complex signal amplitude, a 2 For the second complex signal amplitude, j is the imaginary unit of the complex exponential term. And (3) performing conjugate multiplication on the S1 and the S2 to obtain an interference SAR image: />
Since SAR interferometric phase only records a fractional part of the phase, i.e. its phase value lies between [ -pi, pi).
And S3, marking mining subsidence areas and background areas in any original phase diagram, obtaining and cutting the marked phase diagram, and obtaining a plurality of cutting phase sample diagrams until obtaining a plurality of cutting phase sample diagrams of each original phase diagram.
The mining subsidence area and the background area are marked by using the lableme software, and the marked large-scale image is cut by adopting a sliding window cutting method, and because the cut images do not all contain the mining subsidence area, 1625 small images with the size of 256 multiplied by 256 and each containing the mining subsidence are selected as deep learning samples through manual visual interpretation.
And S4, training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model.
The pytorch is used as a deep learning frame to build an experimental environment, perform model training, parameter adjustment and test, and sample the experimental environment according to 9: the proportion of 1 is randomly divided into a training set and a verification set for training.
The target network model is a model after training of the improved U-net model.
And S5, identifying a target differential interferogram corresponding to the area to be monitored by utilizing the target network model, and obtaining identification results of the mining subsidence area and the background area in the area to be monitored.
Wherein, the area to be monitored is: monitoring areas other than the original monitoring area.
The process of acquiring the target differential interference pattern of the area to be monitored is the same as the process of acquiring the original differential interference pattern of the original area to be monitored.
It should be noted that, when the mining subsidence area and the background area of the area to be monitored are identified, the differential interference pattern of the area to be monitored used is no longer required to be cut. The clipping is used for controlling the same size so that the size of the image is not limited when the computing resource is not wasted to identify the area to be monitored. The image to be identified may be a large image or a small image.
Preferably, the generating, by using a two-track DInSAR technique and based on a first image and a second image of any original monitoring area acquired from the same incident angle at different moments, an original differential interferogram of the any original monitoring area includes:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
Among them, a Digital Elevation Model (DEM) is an essential basis for DInSAR data processing, and this embodiment uses DEM data with a resolution of 30 meters acquired by the us aerospace agency NASA via an SRTM sensor.
In this embodiment, a differential interferogram is generated by adopting a two-track DInSAR technology, and specifically, a process of generating an original differential interferogram of any original monitoring area is shown in fig. 4.
Preferably, the construction process of the improved U-net model is as follows:
and respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model.
Specifically, in this embodiment, a high-Efficiency Channel Attention (ECA) module is added to a coding portion corresponding to each layer from the second layer to the last layer in the original U-Net model, the ECA model belongs to one of the channel attention modules, the ECA module structure is as shown in fig. 5, local cross-channel interaction without dimension reduction is implemented by adopting one-dimensional convolution, and the range of cross-channel interaction is determined by adaptively selecting the size k of the one-dimensional convolution kernel. The ECA module firstly carries out global average pooling operation on the feature graphs of input dimensions [ W, H, C ], after the number of parameters is reduced, the parameters are stored in a matrix with the size of 1 multiplied by C, one-dimensional convolution operation with the size of k of the self-adaptive convolution kernel is carried out to obtain the weights of all channels of the feature graphs, the Sigmoid function normalizes the weights, and then the normalized weights are multiplied with the original feature graphs channel by channel, so that the obtained result is used as the input feature graph of the next stage. As can be seen from comparing fig. 6 and fig. 7, the difference between the structure of the modified U-Net model and the original U-Net model structure is that four ECA modules are added to the corresponding coding portion at the beginning of each of the second layer to the last layer, and the characteristics of the corresponding coding portion before downsampling are readjusted by calculating the weights of the channels.
Preferably, the clipping phase sample map includes: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
before the step S5, the method further includes:
s051, training the original U-net model based on all training sample graphs to obtain an original network model;
s052, sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original model precision of the original network model based on an average cross-correlation formula; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
s053, respectively inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
s054, judging whether the target model precision is larger than the original model precision, obtaining a judging result, and executing step S5 when the judging result is yes.
The original U-Net model and the improved U-Net model are subjected to training, and differences of mining subsidence extraction performance in the differential interference diagram are compared through verification. Model accuracy for the different U-Net models on the validation set is shown in Table 1 (results retain two decimal places).
Table 1: model accuracy of different U-Net models on verification set
From the above table, the improved U-Net model used in this example is superior to the original U-Net model in terms of various indexes, and is improved by 2.54% over IoU (mining subsidence area) compared to the original U-Net model. The improved U-Net model is described as being better for mining subsidence extraction.
According to the technical scheme, the improved U-Net model is used for intelligently identifying the DINSAR interference phase diagram, so that the semantic segmentation precision of the U-Net model is improved, and meanwhile, the identification capability of subsidence areas with different scales is improved.
As shown in fig. 8, a mining subsidence area identification system 200 according to an embodiment of the present invention includes: a preprocessing module 210, a first processing module 220, a first processing module 230, a training module 240, and an identification module 250;
the preprocessing module 210 is configured to: generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated;
the first processing module 220 is configured to: filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain an original phase pattern until the original phase pattern corresponding to each original differential interference pattern is obtained;
the second processing module 230 is configured to: marking mining subsidence areas and background areas in any original phase map, obtaining and cutting the marked phase map to obtain a plurality of cutting phase sample maps until a plurality of cutting phase sample maps of each original phase map are obtained;
the training module 240 is configured to: training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model;
the identification module 250 is configured to: and identifying a target differential interference pattern corresponding to the area to be monitored by utilizing the target network model to obtain identification results of the mining subsidence area and the background area in the area to be monitored.
Preferably, the generating, by using a two-track DInSAR technique and based on a first image and a second image of any original monitoring area acquired from the same incident angle at different moments, an original differential interferogram of the any original monitoring area includes:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
Preferably, the construction process of the improved U-net model is as follows:
and respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model.
Preferably, the clipping phase sample map includes: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
prior to the identification module 250, further comprising: the device comprises a first verification module, a second verification module, a third verification module and a fourth verification module;
the first verification module is used for: training the original U-net model based on all training sample graphs to obtain an original network model;
the second verification module is used for: sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original model precision of the original network model based on an average cross-correlation formula; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
the third verification module is used for: inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
the fourth verification module is used for: and judging whether the target model precision is greater than the original model precision, obtaining a judging result, and calling the identification module 250 when the judging result is yes.
According to the technical scheme, the improved U-Net model is used for intelligently identifying the DINSAR interference phase diagram, so that the semantic segmentation precision of the U-Net model is improved, and meanwhile, the identification capability of subsidence areas with different scales is improved.
The steps for implementing the corresponding functions of the parameters and the modules in the mining subsidence area identification system 200 according to the present embodiment are referred to the parameters and the steps in the embodiment of the mining subsidence area identification method according to the present embodiment, and are not described herein.
The storage medium provided by the embodiment of the invention comprises: the storage medium stores instructions that, when read by a computer, cause the computer to perform steps such as a mining subsidence area identification method, and specific reference may be made to the parameters and steps in the embodiments of a mining subsidence area identification method described above, which are not described herein.
Computer storage media such as: flash disk, mobile hard disk, etc.
The electronic device provided in the embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor executes the computer program to make the computer execute steps of a mining subsidence area identification method, and specific reference may be made to each parameter and step in the embodiment of a mining subsidence area identification method described above, which are not described herein.
Those skilled in the art will appreciate that the present invention may be embodied as methods, apparatus, storage media, and electronic devices.
Thus, the invention may be embodied in the form of: either entirely hardware, entirely software (including firmware, resident software, micro-code, etc.), or entirely software, or a combination of hardware and software, referred to herein generally as a "circuit," module "or" system. Furthermore, in some embodiments, the invention may also be embodied in the form of a computer program product in one or more computer-readable media, which contain computer-readable program code. Any combination of one or more computer readable media may be employed. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.
Claims (6)
1. A mining subsidence area identification method comprising:
s1, generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated;
s2, filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain original phase patterns until the original phase patterns corresponding to each original differential interference pattern are obtained;
s3, marking mining subsidence areas and background areas in any original phase diagram, obtaining and cutting the marked phase diagram to obtain a plurality of cutting phase sample diagrams until a plurality of cutting phase sample diagrams of each original phase diagram are obtained;
s4, training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model;
s5, identifying a target differential interferogram corresponding to the area to be monitored by utilizing the target network model, and obtaining identification results of a mining subsidence area and a background area in the area to be monitored;
the construction process of the improved U-net model comprises the following steps: respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model;
wherein the clipping phase sample map comprises: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
before the step S5, the method further includes:
s051, training the original U-net model based on all training sample graphs to obtain an original network model;
s052, sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original model precision of the original network model based on an average cross-correlation formula; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
s053, respectively inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
s054, judging whether the target model precision is larger than the original model precision, obtaining a judging result, and executing step S5 when the judging result is yes.
2. The mining subsidence area identification method of claim 1, wherein the generating an original differential interferogram of any original monitored area based on a first image and a second image of the any original monitored area at different times acquired from the same incident angle using a two-rail DInSAR technique comprises:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
3. A mining subsidence area identification system comprising: the device comprises a preprocessing module, a first processing module, a second processing module, a training module and an identification module;
the preprocessing module is used for: generating an original differential interference pattern of any original monitoring area by using a two-track method DINSAR technology based on a first image and a second image of the any original monitoring area at different moments acquired from the same incidence angle until the original differential interference pattern of each original monitoring area is generated;
the first processing module is used for: filtering any original differential interference pattern to obtain and convert the filtered differential interference pattern to obtain an original phase pattern until the original phase pattern corresponding to each original differential interference pattern is obtained;
the second processing module is used for: marking mining subsidence areas and background areas in any original phase map, obtaining and cutting the marked phase map to obtain a plurality of cutting phase sample maps until a plurality of cutting phase sample maps of each original phase map are obtained;
the training module is used for: training the improved U-net model based on all clipping phase sample diagrams to obtain a target network model;
the identification module is used for: identifying a target differential interferogram corresponding to the area to be monitored by utilizing the target network model to obtain identification results of a mining subsidence area and a background area in the area to be monitored;
the construction process of the improved U-net model comprises the following steps: respectively adding an efficient channel attention module at the coding part corresponding to each layer from the second layer to the last layer of the original U-Net model, and adjusting the characteristics of the corresponding coding part according to the weight of the channel corresponding to each efficient channel attention module to obtain the improved U-Net model;
wherein the clipping phase sample map comprises: training a sample graph and verifying the sample graph; the mining subsidence area and the background area respectively correspond to a marking category;
before the identification module, further comprising: the device comprises a first verification module, a second verification module, a third verification module and a fourth verification module;
the first verification module is used for: training the original U-net model based on all training sample graphs to obtain an original network model;
the second verification module is used for: sequentially inputting each verification sample graph into the original network model for verification, and obtaining the original of the original network model based on an average cross-correlation formulaModel accuracy; wherein, the average cross ratio formula is:
the tag categories include: a first signature class and a second signature class, the first signature class being the mining subsidence area and the second signature class being the background area, n ii Indicating the number of correctly identified first marker class i, n ji Indicating the number of second marker categories j identified as marker categories i, n cls Representing the number of the marking categories, MIoU representing the average cross ratio;
the third verification module is used for: inputting each verification sample graph into the target network model for verification, and obtaining the target model precision of the target network model based on the average cross-correlation formula;
the fourth verification module is used for: and judging whether the precision of the target model is greater than that of the original model or not, obtaining a judging result, and calling the identification module when the judging result is yes.
4. The mining subsidence area identification system of claim 3, wherein the generating an original differential interferogram for any original monitored area based on first and second images of the any original monitored area at different times acquired from the same angle of incidence using a two-rail DInSAR technique comprises:
and generating an original differential interference pattern of any original monitoring area according to the digital elevation model, the first image and the second image of the any original monitoring area by using a two-rail method DINSAR technology.
5. A storage medium having instructions stored therein which, when read by a computer, cause the computer to perform the mining subsidence area identification method of claim 1 or 2.
6. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program when executed by the processor causes the computer to perform the mining subsidence area identification method of claim 1 or 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210749586.6A CN115331096B (en) | 2022-06-28 | 2022-06-28 | Mining subsidence area identification method, system, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210749586.6A CN115331096B (en) | 2022-06-28 | 2022-06-28 | Mining subsidence area identification method, system, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115331096A CN115331096A (en) | 2022-11-11 |
| CN115331096B true CN115331096B (en) | 2023-04-21 |
Family
ID=83917654
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210749586.6A Active CN115331096B (en) | 2022-06-28 | 2022-06-28 | Mining subsidence area identification method, system, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115331096B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116148853B (en) * | 2023-02-20 | 2024-02-20 | 中国矿业大学 | Mining area subsidence interference phase filtering method and device |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108734211B (en) * | 2018-05-17 | 2019-12-24 | 腾讯科技（深圳）有限公司 | Image processing method and device |
| CN109558859B (en) * | 2018-12-13 | 2020-11-24 | 中国地质大学（武汉） | Mining area distribution information extraction method and system based on DInSAR and DCNN |
| CN111623749B (en) * | 2020-05-19 | 2021-11-16 | 中国铁路设计集团有限公司 | Novel railway goaf boundary extraction method based on D-InSAR |
| CN114167414A (en) * | 2021-11-25 | 2022-03-11 | 中煤航测遥感集团有限公司 | Method for obtaining high-precision surface subsidence basin by unmanned aerial vehicle DEM and InSAR phase fusion |
-
2022
- 2022-06-28 CN CN202210749586.6A patent/CN115331096B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115331096A (en) | 2022-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Piégay et al. | Remotely sensed rivers in the Anthropocene: State of the art and prospects | |
| Ma et al. | Deep learning for geological hazards analysis: Data, models, applications, and opportunities | |
| Gorsevski et al. | Landslide detection and susceptibility mapping using LiDAR and an artificial neural network approach: a case study in the Cuyahoga Valley National Park, Ohio | |
| Wu et al. | Landslide susceptibility assessment using object mapping units, decision tree, and support vector machine models in the Three Gorges of China | |
| Hooshyar et al. | Wet channel network extraction by integrating LiDAR intensity and elevation data | |
| Abdullah et al. | Remote Sensing and Geographic Information System for Fault Segments Mapping a Study from Taiz Area, Yemen. | |
| Shaheen et al. | Data mining applications in hydrocarbon exploration | |
| Hacıefendioğlu et al. | Landslide detection using visualization techniques for deep convolutional neural network models | |
| O'Neil et al. | Effects of LiDAR DEM smoothing and conditioning techniques on a topography‐based wetland identification model | |
| CN115331096B (en) | Mining subsidence area identification method, system, storage medium and electronic equipment | |
| Verbovšek et al. | GIS-assisted classification of litho-geomorphological units using Maximum Likelihood Classification, Vipava Valley, SW Slovenia | |
| Elhag | Consideration of landsat-8 Spectral band combination in typical mediterranean forest classification in Halkidiki, Greece | |
| Mittermeier et al. | Detecting climate change effects on Vb cyclones in a 50‐member single‐model ensemble using machine learning | |
| Virtriana et al. | Machine learning remote sensing using the random forest classifier to detect the building damage caused by the Anak Krakatau Volcano tsunami | |
| Pirasteh et al. | Use of LiDAR-derived DEM and a stream length-gradient index approach to investigation of landslides in Zagros Mountains, Iran | |
| Martínez et al. | Advanced signal recognition methods applied to seismo-volcanic events from Planchon Peteroa Volcanic Complex: Deep Neural Network classifier | |
| Gemusse et al. | Comparing different techniques of satellite imagery classification to mineral mapping pegmatite of Muiane and Naipa: Mozambique | |
| Cagigal et al. | Climate‐based emulator of distant swell trains and local seas approaching a Pacific atoll | |
| Weidner et al. | An algorithm for measuring landslide deformation in terrestrial lidar point clouds using trees | |
| CN109325522B (en) | Heavy industry time sequence heat source region identification algorithm based on improved kmeans | |
| Xu et al. | Building function classification in Nanjing, China, using deep learning | |
| Kaur et al. | A review on natural disaster detection in social media and satellite imagery using machine learning and deep learning | |
| Toulia et al. | The contribution of pattern recognition techniques in geomorphology and geology: the case study of Tinos Island (Cyclades, Aegean, Greece) | |
| Dahl et al. | Magnitude–frequency characteristics and preparatory factors for spatial debris-slide distribution in the northern Faroe Islands | |
| French et al. | Advanced technologies for earthquake risk inventories |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |